A High Resolution Spatially Consistent Global Dataset for CO2 Monitoring

Climate change poses a global threat, affecting both biodiversity and human populations. To implement efficient mitigating strategies, the consistency and accuracy of our monitoring of greenhouse gases at the local level must be improved. We can achieve this with more advanced monitoring instruments or an enhancement of our processing techniques, which will in turn improve data attributes such as spatial or temporal resolutions and accuracy. This paper presents a daily high spatial resolution XCO2 dataset aiming to help monitor atmospheric CO2 concentration on a global scale at a greater level of detail compared with existing datasets. Using a super resolution deep learning model, we increase the resolution of the OCO-2-derived dataset from 0.5° × 0.625° to 0.03° × 0.04° and show that our product maintains the quality of the original dataset while consistently improving the detail of the atmospheric pollution field. We conduct a benchmark that highlights how our dataset outperforms similar products and present a use case of CO2 monitoring at the regional level. In conclusion, this work provides a complementary approach to the area of global continuous dataset reconstruction and focuses on the adjacent problem of improving specific features of existing datasets.